Data Visualization

# Data Visualization

## The grammar of graphics and ggplot2 API

---
class: left

# Today's Agenda

**B. The grammar of graphics and ggplot2 API**
1.	Tidy data
2.  The grammar of graphics
3.	The `{ggplot2}` API 
4.  Questions from the readings
]

**D. How to get help (time permitting)**
]

---

# Introductions

In ~30 seconds, tell us about your experience with R and interest in data science. If you use R (or have used it in the past), what is your biggest data processing, visualization, or data communication challenge?

---
class: left, hide_logo, hide-count

## Setup

.pull-left[
**Option 1**
* Create a folder called `glhlth562` on Box or Dropbox (not for sharing, just for backup)
* Create a subfolder called `glhlth562/materials`
* Create another subfolder called `glhlth562/prep`
* Move your class prep into `glhlth562/prep`
* Download and unzip today's materials into `glhlth562/materials`
]

.pull-right[
**Option 2**
* Clone the repository from https://github.com/ericpgreen/glhlth562
* Create a subfolder called `glhlth562/prep`
* Move your class prep into `glhlth562/prep`
]

---
class: newTopic, hide_logo

# The grammar of graphics and ggplot2 API

---

# Data science life cycle

---

---

---
class: left, hide_logo, hide-count

### No one does this better than the FT's COVID team

---
class: left, hide_logo, hide-count

### John Burn-Murdoch on why he uses {ggplot2}

---

# Tidy data and data wrangling

---

### What is your guess at how the data are structured?

<div class="countdown" id="timer_61ddc340" style="right:0;bottom:0;font-size:1em;" data-warnwhen="0">
<code class="countdown-time"><span class="countdown-digits minutes">01</span><span class="countdown-digits colon">:</span><span class="countdown-digits seconds">00</span></code>
</div>
---

### Wide format by state/county

---
class: left

### Tidy data

According to Hadley Wickham, RStudio's Chief Scientist and creator of many great packages like `{ggplot2}`, [tidy data](https://www.jstatsoft.org/article/view/v059i10) have three characteristics:

1. Each variable forms a column.
2. Each observation forms a row.
3. Each type of observational unit forms a table.

---
class: left

.pull-left[
### Long format
<img src="data:image/png;base64,#img/long.png" width="100%" style="display: block; margin: auto;" />
]

.pull-right[
### State-level data separate
<img src="data:image/png;base64,#img/pop.png" width="100%" style="display: block; margin: auto;" />
]

---

### There is (usually) no free lunch

.pull-left[
Data are rarely ready to plot right out of the box. We'll skip the import and wrangling details for now, but take note (right) of what's required in this case.
]

.verysmall[
.pull-right[
```
pop <- read_csv("https://raw.githubusercontent.com/CSSEGISandData/COVID-19/master/csse_covid_19_data/csse_covid_19_time_series/time_series_covid19_deaths_US.csv") %>%
  # keep only the columns we need
    select(Province_State, Population) %>%
  # keep only the rows we need
    filter(Province_State=="North Carolina" | Province_State=="Arizona") %>%
  # aggregate population
    group_by(Province_State) %>%
    summarize(Population = sum(Population))

covid <- read_csv("https://raw.githubusercontent.com/CSSEGISandData/COVID-19/master/csse_covid_19_data/csse_covid_19_time_series/time_series_covid19_deaths_US.csv") %>%
  # keep only the columns we need
    select(Province_State, Population, contains("/")) %>%
  # keep only the rows we need
    filter(Province_State=="North Carolina" | Province_State=="Arizona") %>%
  # make long
    pivot_longer(cols = c(-Province_State, -Population),
                 names_to = "date",
                 values_to = "totalDeaths") %>%
  # convert date to proper date
    mutate(date = mdy(date)) %>%
  # aggregate by state and date
    group_by(Province_State, date) %>%
    summarize(totalDeaths = sum(totalDeaths)) %>%
  # calculate daily deaths
  # end up with some negative values, likely due to data revisions
    group_by(Province_State) %>%
    mutate(dailyDeaths = c(totalDeaths[1], diff(totalDeaths))) %>%
  # make sure series is complete
    ungroup() %>%
    complete(Province_State, date, fill = list(0)) %>%
  # keep only dates in range
    filter(date >= "2020-08-20" & date <= "2021-07-12") %>%
  # calculate rolling mean
    group_by(Province_State) %>%
    mutate(deaths_roll7 = zoo::rollmean(dailyDeaths, k = 7, fill = NA)) %>%
    filter(!is.na(deaths_roll7)) %>%
    filter(date >= "2020-09-01") %>%
  # merge in pop data and normalize
    left_join(pop, by="Province_State") %>%
    mutate(deaths_roll7_100k = (deaths_roll7/Population)*100000) %>%
  # label the last entry for each state (for the plotting step)
    mutate(label = if_else(date == max(date), 
                           Province_State, 
                           NA_character_)) %>%
    select(Province_State, date, deaths_roll7_100k, label)
``` 
] 
]

---

# Grammar of graphics

---

### What are the key plot elements?

---

### What are the key plot elements?

.pull-left[
<img src="data:image/png;base64,#img/ft2.png" width="100%" style="display: block; margin: auto;" />
]

---

### What are the key plot elements?

.pull-left[
<img src="data:image/png;base64,#img/ft2.png" width="100%" style="display: block; margin: auto;" />
]

.pull-right[
* Lines by group with group labels
* Axis labels
* Headline title
* Down to business subtitle
* Logo
* Caption with data source
* FT theme
* Tool tip
]

---
class: left

# Grammar of graphics

`{ggplot2}` implements the Leland Wilkinson's "grammar of graphics". Wickham (2010) describes the grammar of graphics as:

"...a tool that enables us to concisely describe the components of a graphic. Such a grammar allows us to move beyond named graphics (e.g., the "scatterplot") and gain insight into the deep structure that underlies statistical graphics."

---
class: left

<span style="font-size: 50%;">Source: Thomas Lin Pederson. Amazing workshop on `ggplot2`: https://github.com/thomasp85/ggplot2_workshop</span>

---
class: left

<span style="font-size: 50%;">Source: Thomas de Beus, https://tinyurl.com/sf8zude</span>

---
class: newTopicSub, hide_logo

# {ggplot2} API

---

### Let's use this example to explore the API

---

### Open RStudio

.pull-left[
* Find and open the `dataviz1_template.Rmd` file in `materials/dataviz1`. 
* Install packages in `setup` chunk
* Scroll to `canvas` chunk
* Press the arrow pointing down

(We'll dig into the details of RMarkdown files like this later)
]

.pull-right[
<img src="data:image/png;base64,#img/open.png" width="100%" style="display: block; margin: auto;" />
]

---
class: left

### `glimpse()` the processed data

The data are stored in an object called `covid`. In addition to running `glimpse(covid)`, you can also click on `covid` in the Environment to see a 'spreadsheet' view.

```
## Rows: 624
## Columns: 6
## $ X1                <dbl> 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 1…
## $ Province_State    <chr> "Arizona", "Arizona", "Arizona", "Arizona", "Arizona…
## $ date              <date> 2020-09-01, 2020-09-02, 2020-09-03, 2020-09-04, 202…
## $ deaths_roll7      <dbl> 27.57143, 28.57143, 27.28571, 27.14286, 25.28571, 26…
## $ deaths_roll7_100k <dbl> 0.3787952, 0.3925339, 0.3748698, 0.3729072, 0.347392…
## $ label             <chr> NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, …
```

---

### The `ggplot()` function

`ggplot()` tells R that we want to plot something, but we haven't specified any data or told R how the data map to this plot canvas.

```r
# see chunk 'canvas'
ggplot()
```
]

<img src="data:image/png;base64,#dataviz1_deck_files/figure-html/canvas_plot-1.png" width="100%" />
]

]

---

### Map data to the canvas

The first step is to tell `ggplot()` what data you want to plot and specify how the data maps onto the canvas. Which variable maps to the `x` axis? To the `y` axis?

```r
# see chunk 'mapping'
ggplot(covid, aes(x = ____, y = ____))
```
]

```r
ggplot(covid, aes(x = date, y = deaths_roll7_100k))
```
]

<img src="data:image/png;base64,#dataviz1_deck_files/figure-html/mapping_plot-1.png" width="100%" />
]

]

---

### Add a `geom_`

---

### Add a `geom_`

Our plot has appropriate axes now that we've mapped the data, but the canvas is blank because we've not told R how to represent the data. Our inspiration plot uses lines, so we will too.

```r
# see chunk 'geom'
ggplot(covid, aes(x = date, y = deaths_roll7_100k)) +
  geom_line()
```
]

<img src="data:image/png;base64,#dataviz1_deck_files/figure-html/geom_plot-1.png" width="100%" />
]

]
---

### Specify the grouping

This does not look right because we did not tell R that the data are grouped. `covid` is a 'long' dataset with two series: Arizona and North Carolina. Which variable contains this grouping information?

```r
# see chunk 'grouping'
ggplot(covid, aes(x = date, y = deaths_roll7_100k,
                  group = ____)) +
  geom_line()
```
]

```r
ggplot(covid, aes(x = date, y = deaths_roll7_100k,
                  group = Province_State)) +
  geom_line()
```
]

<img src="data:image/png;base64,#dataviz1_deck_files/figure-html/grouping_plot-1.png" width="100%" />
]

]

---
class: left

### Specify the grouping color

We need to add color to identify each state's line. One point of friction for new users is deciding whether to use `color` or `fill`. Remember this: Lines and points use `color`.

```r
# see chunk 'addColor'
ggplot(covid, aes(x = date, y = deaths_roll7_100k,
                  color = Province_State)) +
  geom_line()
```

Rather than specifying `group` explicitly, we can simplify to `color`, which implicitly tells R that we're coloring by group.

]

<img src="data:image/png;base64,#dataviz1_deck_files/figure-html/addColor_plot-1.png" width="100%" />
]

]

---

### Modify the style

We'll dig into plot refinements later. For now, just take note of the layering approach.

```r
# see chunk 'changeColors'
ggplot(covid,                                     # data
       aes(x = date, y = deaths_roll7_100k,       # mapping
           color = Province_State)) +
       geom_line() +                              # geom
       scale_color_manual(values = c(             # theme
                                     "#eb5e8d",     # pink 
                                     "#0f5599"      # blue
                                     )) +
      ft_theme()                                  
```

]

<img src="data:image/png;base64,#dataviz1_deck_files/figure-html/changeColors_plot-1.png" width="70%" />
]
]
---

<span style="font-size: 50%;">Source: Thomas Lin Pederson. Amazing workshop on `ggplot2`: https://github.com/thomasp85/ggplot2_workshop</span>

---

# Credits

Deck by Eric Green ([@ericpgreen](https://twitter.com/ericpgreen)), licensed under Creative Commons Attribution [CC BY-SA 4.0](https://creativecommons.org/licenses/by-sa/4.0/)

* {[`xaringan`](https://github.com/yihui/xaringan)} for slides with help from {[`xaringanExtra`](https://github.com/gadenbuie/xaringanExtra)} 
* Data science lifecycle, [Data Science in a Box](https://datasciencebox.org/)
* [JHU CSSE COVID-19 data](https://github.com/CSSEGISandData/COVID-19)